Three-Dimensional Semantic Segmentation of Diabetic Retinopathy Lesions and Grading Using Transfer Learning

Abstract

Diabetic retinopathy (DR) is a drastic disease. DR embarks on vision impairment when it is left undetected. In this article, learning-based techniques are presented for the segmentation and classification of DR lesions. The pre-trained Xception model is utilized for deep feature extraction in the segmentation phase. The extracted features are fed to Deeplabv3 for semantic segmentation. For the training of the segmentation model, an experiment is performed for the selection of the optimal hyperparameters that provided effective segmentation results in the testing phase. The multi-classification model is developed for feature extraction using the fully connected (FC) MatMul layer of efficient-net-b0 and pool-10 of the squeeze-net. The extracted features from both models are fused serially, having the dimension of N × 2020, amidst the best N × 1032 features chosen by applying the marine predictor algorithm (MPA). The multi-classification of the DR lesions into grades 0, 1, 2, and 3 is performed using neural network and KNN classifiers. The proposed method performance is validated on open access datasets such as DIARETDB1, e-ophtha-EX, IDRiD, and Messidor. The obtained results are better compared to those of the latest published works.

Level-set based adaptive-active contour segmentation technique with long short-term memory for diabetic retinopathy classification

“…The comparison of proposed LBACS-LSTM with the existing approaches such as three-dimensional semantic model ( Jebaseeli et al, 2019 ), KNN ( Rachapudi et al, 2023 ), Computer-aided method ( Shaukat et al, 2022 ), CTSA-SAE ( Dayana and Emmanuel, 2022a ; Dayana and Emmanuel, 2022b ), DS-KL ( Mondal et al, 2023 ), ESOA optimized hybrid RCNN-BiGRU ( Alajlan and Razaque, 2023 ) and Adaptive CNN ( Math and Fatima, 2021 ) are described in this section. Table 10 presents the comparative analysis for IDRiD dataset and Table 11 depicts the comparative analysis of DIARETDB 1 dataset.…”

“…The results from Table 10 and Table 11 show that the proposed LBACS-LSTM achieves better performance in overall metrics when compared with the existing three-dimensional semantic models ( Jebaseeli et al, 2019 ), KNN ( Rachapudi et al, 2023 ), Computer-aided method ( Shaukat et al, 2022 ), CTSA-SAE ( Dayana and Emmanuel, 2022b ), DS-KL ( Mondal et al, 2023 ) and ESOA optimized hybrid RCNN-BiGRU ( Alajlan and Razaque, 2023 ) and Adaptive CNN ( Math and Fatima, 2021 ). The accuracy of the proposed method for the IDRiD dataset is 99.43% and 97.39% for the DIARETDB 1 dataset.…”

“… Shaukat et al (2022) introduced a three-dimensional semantic model to segment the diabetic retinopathy lesions. The pre-processed input data was fed into the pre-trained Xception model, and the extracted features were segmented with the help of Deeplabv3.…”

An Improved Residual U-Net for Segmentation of Multiple Structures in Fundus Images

“…Learning-based methods for segmenting and categorizing DR lesions are provided in [15]. Deep feature extraction is carried out during the segmentation level utilizing the trained Xception system.…”

Prediction of Alzheimer's disease stages based on ResNet-Self-attention architecture with Bayesian optimization and best features selection

“…Classification is the bottommost step in which each object accredits a label, in either a supervised or unsupervised ML technique (Bondi et al, 2017 ). Deep learning (DL) (Shaukat et al, 2022 ) is a subtype of machine learning that falls under the umbrella of artificial intelligence, but DL is way more vigorous and flexible in comparison with ML (Fabrizio et al, 2021 ). Techniques such as shallow CNN (Marwa et al, 2023 ), DNN (Hazarika et al, 2023 ), MultiAz-Net (Ismail et al, 2023 ), hybridized DL method (Hashmi, 2024 ), and RVFL (Goel et al, 2023 ) have been used in recent years, but these techniques yield low accuracy as compared to our proposed model (Shamrat et al, 2023 ).…”